Lane-Changing Behavior Recognition in the Connected Vehicle Environment

Abstract

AbstractTraditional lane-changing models are mostly based on mathematical and physical models that do not adequately take into account the uncertainty and randomness of the driver behavior in a complex road environment, and the data for training models are often generated from driving simulators rather than from vehicles traveling under real road conditions. In this paper, a neural network model is proposed to identify a driver’s lane changing behavior based on real-time driving data collected in the connected vehicle environment. An in-depth analysis of the lane-changing process is conducted, dividing the process into different stages and identifying the factors that characterize the lane-changing behavior. Research data is extracted from the Safety Pilot Model Development (SPMD) project. Feedforward Neural Network (FNN) is selected to construct the model given its good ability for nonlinear fitting. The model is trained using SPMD data to determine the optimal network structure. Applying the trained model to identify the lane-changing behavior, the corresponding recognition rate under the optimal time window reaches 90.7%. The curve of Receiver Operating Characteristic (ROC) shows that the FNN model can satisfy the requirements for the identification of lane-changing vehicles. The findings of this paper may contribute to the safety of vehicle lane changing in the connected vehicle environment and reduce traffic crashes caused by lane changing through advance warnings.KeywordsLane-changing behaviorSPMD dataFeedforward neural networkConnected vehicle